software-realtime
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The scripts
scripts/check_ws_smoke.pyandscripts/ws_smoke_test.pyinitiate network connections to arbitrary WebSocket URLs. While intended for testing, this capability can be used for data transfer. - Evidence: Scripts use
socket.create_connectionand libraries likewebsocketsorwebsocket-clientto establish external connections. - [INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection via network responses from untrusted servers.
- Ingestion points: WebSocket server responses in diagnostic scripts enter the agent context.
- Boundary markers: No delimiters or instructions are present for processing external data.
- Capability inventory: Network access and message processing.
- Sanitization: None.
Audit Metadata